Convergent Manufacturing: Where AI, Materials and Automation Meet

Awesome. Great. Yeah. Hello everyone. Thank you so much for inviting me to this podcast. I'm very excited to talk about manufacturing in general and particularly about our research.

Well, in a very unique way, I can tell you. So the most important thing is that, you know, because of my industry background, whenever I'm trying to define a new research project, I try to look at that from the perspective of industry, meaning that, okay, what problem are we solving for industry, what is the current practice in the industry related to that topic, and what are the challenges and shortcomings of that current practice. Also, I'll try to kind of put my entrepreneurship hat on and think about the value of that problem. Okay? Because whatever solution that we propose should be proportionate to the effort needed to realize that solution. Okay? And that should be kind of at the same scale of the problem that we are solving. In general, I can tell you that this combined industry-entrepreneurship background helped me to switch my approach from solution-based innovation into problem-based innovation, meaning that, you know, this is very common in academia that we try to fit our research into a particular problem, you know, like we have a solution, we try to fit it somewhere, okay, whether it can fix this issue or that issue. But I try to stay away from that, you know, and saying that, okay, what is the real problem here, and what is the best solution for this problem, and what are the scientific challenges ahead of this solution that I can tackle through my research, you know, and then try to define my research on those basically research gaps and on those challenges. So it's a bit kind of switching from solution-based innovation into problem-based innovation. So that was the main, I feel, the influence of my background into my academic life.

Exactly. Yeah, that's correct. You know, I'm trying to look at that from different angles and from different perspectives, you know, seeing that, you know, how much this solution works for industry. You know, to give you a very good example, you know, we recently came up with a kind of a low-cost tooling solution for prototyping, for low-volume production, for proof-of-concept production. I can quickly bring the sample. I don't know whether it's possible or not. Let's do it. We can figure it out later. So in this solution, okay, we have a 3D-printed insert with a sheet metal lamination over there. Okay. So we get the quality of metal tools with the cost and lead time of, you know, FDM 3D printing. Okay. So it's very groundbreaking, and, you know, tooling cost is the main barrier for industry to adopt new technology, to develop new products. Okay. So we are tackling a very good, basically a very important problem in the industry. But through this project, we realized that in order to have an accurate part at the end, we need to spend lots of engineering time to design this low-cost tool. So then I realized that from the perspective of my entrepreneurship background, if we need to spend like 100 hours of engineering to realize this low-cost tool, you know, it's better to go get a piece of metal and just machine it. You know, although we reduce the cost of the material and cost of the manufacturing with this solution, but we are significantly increasing the cost of the engineering. So at the end, this is not going to be a low-cost solution. So that's why we pivoted our research into how we can reduce the engineering effort for such kind of low-cost tooling by leveraging more AI and more basically low-cost but high-fidelity modeling techniques. So this was a good example of that. You know, in a normal scenario, we would say that, okay, this is a groundbreaking, awesome, you know, kind of solution, but industry wouldn't adopt it because, okay, it needs a significant amount of engineering work, okay, it's not economical at the end, okay. But now we are pivoting, okay, from that direction, saying that, okay, this is a great idea, but now let's focus on how we can reduce the engineering cost.

Exactly. We wanted to see how we can replace basically high-computational-cost simulation with reduced-order surrogate models that can quickly predict the performance of these tools, and we can even use it for inverse engineering and inverse design, quickly finding the optimum configuration of these tools. For example, what is the material for the insert? What is the material for the sheet metal lamination, thickness of them, you know, the infill design, and the process parameters to get the, you know, right product at the end, you know, like it's a complex manufacturing chain. But those digital twins, those computationally light digital twins, can help us to quickly, you know, go and find optimum or near-optimum solutions. Okay, instead of going into multiple back-and-forth simulations, trials, you know, to realize that, okay, what would be the right design configuration for that.

That's a great question. You know, to me, we are going to see more interdisciplinary R&D projects in the future. As a result, we will see more revolutionary products, product design, and significant improvement in how we can make our products, you know, rather than incremental improvement. The reason is that, you know, these new AI tools can help us to more efficiently expand our knowledge, okay, and basically conceptualize such kind of interdisciplinary or multidisciplinary ideas, and even more efficiently execute those ideas and realize those ideas. You know, in the past, when we wanted to come up with such kind of interdisciplinary ideas, we had to assemble a group of researchers, okay, and spend lots of time of discussion and going through, you know, back-and-forth kind of communication to come up with a coherent interdisciplinary research idea. Okay. And then when we wanted to, you know, execute that research and we wanted to realize that idea, we needed, you know, substantial funding because it involves multiple groups, you know, from different disciplines and multiple entities. Okay. But right now, these new AI tools, especially these, you know, LLM tools, can help you, you know, to much faster, you know, learn about a new field, collect, you know, literature, collect all the data related to that, summarize them, convert that to the language that you can understand, and walk you step by step through that field of science. So now one researcher, for example, myself in manufacturing, if I need to know anything about computer science for a particular application, I can make myself up to date just to at least conceptualize that idea. Okay. Then later I need to work maybe with some other party to realize it. You know, if it's a really big project, but even if it's some small project, even I can get help, you know, to the level that I need and kind of realize the initial version of that product. So that is something that I feel will be the most influential effect of AI in the future, particularly in the area of manufacturing.

Yeah, convergent manufacturing is a relatively new field of manufacturing science. In this field, we explore how we can synergistically combine heterogeneous materials, manufacturing processes, and advanced technologies such as robotics, AI, and IoT solutions to find an optimum design and manufacturing solution for a given set of requirements. In simpler terms, I can tell you that convergent manufacturing is a visionary field in which we systematically explore the interdisciplinary opportunities that I mentioned previously to find an optimum or near-optimum solution for a given problem. Okay. So kind of that is very important because it's a very broad design space and very broad range of opportunities when you consider, you know, all combinations of the materials and manufacturing processes and even some advanced technologies. So now laser-pointing the optimum solution, it's a very difficult, you know, problem to solve. Okay. So that's something, you know, we focus on that, okay, how we can efficiently find that optimal or near-optimal. Okay, so that's kind of the focus of convergent manufacturing. I wanted to kind of explain this convergent manufacturing on an example, you know, for your audience to better understand. For example, imagine we are trying to design a battery enclosure for an EV. Okay. In a conventional approach, we say that, okay, battery enclosure needs to have some structural performance. Okay. Meaning that in terms of the bending strength, you know, compression strength, you know, penetration in terms of any external object, and so on and so on. So at first, okay, let's design a sheet metal or composite box to fulfill that structural requirement. Okay. So we need to have a structural enclosure, and then we go to the next set of requirements saying that, okay, this enclosure also needs to dissipate heat generated by battery cells. Okay. Now let's add a cooling system to this enclosure. And then now we see, okay, we need a feedback system. Okay. We wanted to have a thermal control system. So now let's assemble some sensors in some location of the enclosure and so on. The result is a product that fulfills all the requirements. Okay, no doubt about that. But most probably it's going to be a bulky, you know, product and needs a set of different, you know, manufacturing stages and assembly steps which require lots of capital cost, material cost, and manufacturing cost. Okay, because, you know, we consider all these requirements individually and assemble them together as a kind of an assembly. Okay. But in convergent manufacturing, our approach is different. We combine all the requirements at the earliest stage of the design and try to see how we can combine different materials and manufacturing processes to develop a multi-material enclosure with integrated additional functionality such as cooling, sensing, electrical, and even safety features, and how we can produce this multi-functional enclosure with a hybrid manufacturing system. Basically combining multiple manufacturing processes into one manufacturing cell, physically integrating them using automation and robotic systems, and virtually orchestrating the whole technology using supervisory control and data acquisition systems or any other digital tools. The result will be a very concise product set, something like this. You know, one of our recent basically project results, as you see here. This is a battery enclosure that has the structural functionality with a composite sheet on the back and has cooling functionality with the embedded kind of lattice core structure at this region, and also a sheet metal that takes care of the heat dissipation to the cooling channel. So all these things happen in one single hybrid molding operation, you know, like putting the sheet metals, the core, putting the composite sheet, the lattice core, sheet metal, all into one single specialized tool, and then molding all of them together into one basically single consolidated product. So this is a kind of a brief, to me it was a brief explanation about what is convergent manufacturing.

Perfectly. So efficiency at different levels, you know, when you integrate multiple processes into one singular operation, you significantly reduce capital cost because instead of having multiple separate equipment, each of them needs separate tooling and separate operators and also lots of production planning effort behind that. So these are all lots of cost, okay, capital cost, manufacturing cost, and sometimes energy cost, you know, if you wanted to make like this part separately and then put it into another process to heat it and bond it together. So now, so you first heat that part in the first process, cool it down, and now again another process, heat it again up, make it ready, for example, for a subsequent overmolding process. Okay. So it's lots of efficiency plus efficiency in terms of using the material because we use material where we need that property. For example, here we only need sheet metal in that region to dissipate heat generated by battery cells. So the rest doesn't need to be a sheet metal. Okay? So you can save the cost of that. A structural part only needs to be at the back of that. So you don't need a structural part, like continuous fiber composite, in the rest of the enclosure. So this is another way of basically cutting the cost and reaching to the highest level of efficiency possible.